aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/staging/lustre/lustre/ptlrpc/pinger.c
blob: 5dec771d70eee8c08a6bc0b787f7f116ddd6906d (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
/*
 * GPL HEADER START
 *
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 only,
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License version 2 for more details (a copy is included
 * in the LICENSE file that accompanied this code).
 *
 * You should have received a copy of the GNU General Public License
 * version 2 along with this program; If not, see
 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
 *
 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
 * CA 95054 USA or visit www.sun.com if you need additional information or
 * have any questions.
 *
 * GPL HEADER END
 */
/*
 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
 * Use is subject to license terms.
 *
 * Copyright (c) 2011, 2012, Intel Corporation.
 */
/*
 * This file is part of Lustre, http://www.lustre.org/
 * Lustre is a trademark of Sun Microsystems, Inc.
 *
 * lustre/ptlrpc/pinger.c
 *
 * Portal-RPC reconnection and replay operations, for use in recovery.
 */

#define DEBUG_SUBSYSTEM S_RPC

#include <obd_support.h>
#include <obd_class.h>
#include "ptlrpc_internal.h"

static int suppress_pings;
CFS_MODULE_PARM(suppress_pings, "i", int, 0644, "Suppress pings");

struct mutex pinger_mutex;
static LIST_HEAD(pinger_imports);
static struct list_head timeout_list = LIST_HEAD_INIT(timeout_list);

int ptlrpc_pinger_suppress_pings(void)
{
	return suppress_pings;
}
EXPORT_SYMBOL(ptlrpc_pinger_suppress_pings);

struct ptlrpc_request *
ptlrpc_prep_ping(struct obd_import *imp)
{
	struct ptlrpc_request *req;

	req = ptlrpc_request_alloc_pack(imp, &RQF_OBD_PING,
					LUSTRE_OBD_VERSION, OBD_PING);
	if (req) {
		ptlrpc_request_set_replen(req);
		req->rq_no_resend = req->rq_no_delay = 1;
	}
	return req;
}

int ptlrpc_obd_ping(struct obd_device *obd)
{
	int rc;
	struct ptlrpc_request *req;

	req = ptlrpc_prep_ping(obd->u.cli.cl_import);
	if (req == NULL)
		return -ENOMEM;

	req->rq_send_state = LUSTRE_IMP_FULL;

	rc = ptlrpc_queue_wait(req);

	ptlrpc_req_finished(req);

	return rc;
}
EXPORT_SYMBOL(ptlrpc_obd_ping);

int ptlrpc_ping(struct obd_import *imp)
{
	struct ptlrpc_request *req;

	req = ptlrpc_prep_ping(imp);
	if (req == NULL) {
		CERROR("OOM trying to ping %s->%s\n",
		       imp->imp_obd->obd_uuid.uuid,
		       obd2cli_tgt(imp->imp_obd));
		return -ENOMEM;
	}

	DEBUG_REQ(D_INFO, req, "pinging %s->%s",
		  imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd));
	ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);

	return 0;
}

void ptlrpc_update_next_ping(struct obd_import *imp, int soon)
{
	int time = soon ? PING_INTERVAL_SHORT : PING_INTERVAL;
	if (imp->imp_state == LUSTRE_IMP_DISCON) {
		int dtime = max_t(int, CONNECTION_SWITCH_MIN,
				  AT_OFF ? 0 :
				  at_get(&imp->imp_at.iat_net_latency));
		time = min(time, dtime);
	}
	imp->imp_next_ping = cfs_time_shift(time);
}

void ptlrpc_ping_import_soon(struct obd_import *imp)
{
	imp->imp_next_ping = cfs_time_current();
}

static inline int imp_is_deactive(struct obd_import *imp)
{
	return (imp->imp_deactive ||
		OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_IMP_DEACTIVE));
}

static inline int ptlrpc_next_reconnect(struct obd_import *imp)
{
	if (imp->imp_server_timeout)
		return cfs_time_shift(obd_timeout / 2);
	else
		return cfs_time_shift(obd_timeout);
}

static atomic_t suspend_timeouts = ATOMIC_INIT(0);
static cfs_time_t suspend_wakeup_time = 0;

cfs_duration_t pinger_check_timeout(cfs_time_t time)
{
	struct timeout_item *item;
	cfs_time_t timeout = PING_INTERVAL;

	/* The timeout list is a increase order sorted list */
	mutex_lock(&pinger_mutex);
	list_for_each_entry(item, &timeout_list, ti_chain) {
		int ti_timeout = item->ti_timeout;
		if (timeout > ti_timeout)
			timeout = ti_timeout;
		break;
	}
	mutex_unlock(&pinger_mutex);

	return cfs_time_sub(cfs_time_add(time, cfs_time_seconds(timeout)),
					 cfs_time_current());
}

static wait_queue_head_t suspend_timeouts_waitq;

cfs_time_t ptlrpc_suspend_wakeup_time(void)
{
	return suspend_wakeup_time;
}

void ptlrpc_deactivate_timeouts(struct obd_import *imp)
{
	/*XXX: disabled for now, will be replaced by adaptive timeouts */
#if 0
	if (imp->imp_no_timeout)
		return;
	imp->imp_no_timeout = 1;
	atomic_inc(&suspend_timeouts);
	CDEBUG(D_HA|D_WARNING, "deactivate timeouts %u\n",
	       atomic_read(&suspend_timeouts));
#endif
}

void ptlrpc_activate_timeouts(struct obd_import *imp)
{
	/*XXX: disabled for now, will be replaced by adaptive timeouts */
#if 0
	if (!imp->imp_no_timeout)
		return;
	imp->imp_no_timeout = 0;
	LASSERT(atomic_read(&suspend_timeouts) > 0);
	if (atomic_dec_and_test(&suspend_timeouts)) {
		suspend_wakeup_time = cfs_time_current();
		wake_up(&suspend_timeouts_waitq);
	}
	CDEBUG(D_HA|D_WARNING, "activate timeouts %u\n",
	       atomic_read(&suspend_timeouts));
#endif
}

int ptlrpc_check_suspend(void)
{
	if (atomic_read(&suspend_timeouts))
		return 1;
	return 0;
}

int ptlrpc_check_and_wait_suspend(struct ptlrpc_request *req)
{
	struct l_wait_info lwi;

	if (atomic_read(&suspend_timeouts)) {
		DEBUG_REQ(D_NET, req, "-- suspend %d regular timeout",
			  atomic_read(&suspend_timeouts));
		lwi = LWI_INTR(NULL, NULL);
		l_wait_event(suspend_timeouts_waitq,
			     atomic_read(&suspend_timeouts) == 0, &lwi);
		DEBUG_REQ(D_NET, req, "-- recharge regular timeout");
		return 1;
	}
	return 0;
}


static bool ir_up;

void ptlrpc_pinger_ir_up(void)
{
	CDEBUG(D_HA, "IR up\n");
	ir_up = true;
}
EXPORT_SYMBOL(ptlrpc_pinger_ir_up);

void ptlrpc_pinger_ir_down(void)
{
	CDEBUG(D_HA, "IR down\n");
	ir_up = false;
}
EXPORT_SYMBOL(ptlrpc_pinger_ir_down);

static void ptlrpc_pinger_process_import(struct obd_import *imp,
					 unsigned long this_ping)
{
	int level;
	int force;
	int force_next;
	int suppress;

	spin_lock(&imp->imp_lock);

	level = imp->imp_state;
	force = imp->imp_force_verify;
	force_next = imp->imp_force_next_verify;
	/*
	 * This will be used below only if the import is "FULL".
	 */
	suppress = ir_up && OCD_HAS_FLAG(&imp->imp_connect_data, PINGLESS);

	imp->imp_force_verify = 0;

	if (cfs_time_aftereq(imp->imp_next_ping - 5 * CFS_TICK, this_ping) &&
	    !force) {
		spin_unlock(&imp->imp_lock);
		return;
	}

	imp->imp_force_next_verify = 0;

	spin_unlock(&imp->imp_lock);

	CDEBUG(level == LUSTRE_IMP_FULL ? D_INFO : D_HA, "%s->%s: level %s/%u "
	       "force %u force_next %u deactive %u pingable %u suppress %u\n",
	       imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd),
	       ptlrpc_import_state_name(level), level, force, force_next,
	       imp->imp_deactive, imp->imp_pingable, suppress);

	if (level == LUSTRE_IMP_DISCON && !imp_is_deactive(imp)) {
		/* wait for a while before trying recovery again */
		imp->imp_next_ping = ptlrpc_next_reconnect(imp);
		if (!imp->imp_no_pinger_recover)
			ptlrpc_initiate_recovery(imp);
	} else if (level != LUSTRE_IMP_FULL ||
		   imp->imp_obd->obd_no_recov ||
		   imp_is_deactive(imp)) {
		CDEBUG(D_HA, "%s->%s: not pinging (in recovery "
		       "or recovery disabled: %s)\n",
		       imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd),
		       ptlrpc_import_state_name(level));
	} else if ((imp->imp_pingable && !suppress) || force_next || force) {
		ptlrpc_ping(imp);
	}
}

static int ptlrpc_pinger_main(void *arg)
{
	struct ptlrpc_thread *thread = (struct ptlrpc_thread *)arg;

	/* Record that the thread is running */
	thread_set_flags(thread, SVC_RUNNING);
	wake_up(&thread->t_ctl_waitq);

	/* And now, loop forever, pinging as needed. */
	while (1) {
		cfs_time_t this_ping = cfs_time_current();
		struct l_wait_info lwi;
		cfs_duration_t time_to_next_wake;
		struct timeout_item *item;
		struct list_head *iter;

		mutex_lock(&pinger_mutex);
		list_for_each_entry(item, &timeout_list, ti_chain) {
			item->ti_cb(item, item->ti_cb_data);
		}
		list_for_each(iter, &pinger_imports) {
			struct obd_import *imp =
				list_entry(iter, struct obd_import,
					       imp_pinger_chain);

			ptlrpc_pinger_process_import(imp, this_ping);
			/* obd_timeout might have changed */
			if (imp->imp_pingable && imp->imp_next_ping &&
			    cfs_time_after(imp->imp_next_ping,
					   cfs_time_add(this_ping,
							cfs_time_seconds(PING_INTERVAL))))
				ptlrpc_update_next_ping(imp, 0);
		}
		mutex_unlock(&pinger_mutex);
		/* update memory usage info */
		obd_update_maxusage();

		/* Wait until the next ping time, or until we're stopped. */
		time_to_next_wake = pinger_check_timeout(this_ping);
		/* The ping sent by ptlrpc_send_rpc may get sent out
		   say .01 second after this.
		   ptlrpc_pinger_sending_on_import will then set the
		   next ping time to next_ping + .01 sec, which means
		   we will SKIP the next ping at next_ping, and the
		   ping will get sent 2 timeouts from now!  Beware. */
		CDEBUG(D_INFO, "next wakeup in "CFS_DURATION_T" ("
		       CFS_TIME_T")\n", time_to_next_wake,
		       cfs_time_add(this_ping,cfs_time_seconds(PING_INTERVAL)));
		if (time_to_next_wake > 0) {
			lwi = LWI_TIMEOUT(max_t(cfs_duration_t,
						time_to_next_wake,
						cfs_time_seconds(1)),
					  NULL, NULL);
			l_wait_event(thread->t_ctl_waitq,
				     thread_is_stopping(thread) ||
				     thread_is_event(thread),
				     &lwi);
			if (thread_test_and_clear_flags(thread, SVC_STOPPING)) {
				break;
			} else {
				/* woken after adding import to reset timer */
				thread_test_and_clear_flags(thread, SVC_EVENT);
			}
		}
	}

	thread_set_flags(thread, SVC_STOPPED);
	wake_up(&thread->t_ctl_waitq);

	CDEBUG(D_NET, "pinger thread exiting, process %d\n", current_pid());
	return 0;
}

static struct ptlrpc_thread pinger_thread;

int ptlrpc_start_pinger(void)
{
	struct l_wait_info lwi = { 0 };
	int rc;

	if (!thread_is_init(&pinger_thread) &&
	    !thread_is_stopped(&pinger_thread))
		return -EALREADY;

	init_waitqueue_head(&pinger_thread.t_ctl_waitq);
	init_waitqueue_head(&suspend_timeouts_waitq);

	strcpy(pinger_thread.t_name, "ll_ping");

	/* CLONE_VM and CLONE_FILES just avoid a needless copy, because we
	 * just drop the VM and FILES in cfs_daemonize_ctxt() right away. */
	rc = PTR_ERR(kthread_run(ptlrpc_pinger_main, &pinger_thread,
				 "%s", pinger_thread.t_name));
	if (IS_ERR_VALUE(rc)) {
		CERROR("cannot start thread: %d\n", rc);
		return rc;
	}
	l_wait_event(pinger_thread.t_ctl_waitq,
		     thread_is_running(&pinger_thread), &lwi);

	if (suppress_pings)
		CWARN("Pings will be suppressed at the request of the "
		      "administrator.  The configuration shall meet the "
		      "additional requirements described in the manual.  "
		      "(Search for the \"suppress_pings\" kernel module "
		      "parameter.)\n");

	return 0;
}

int ptlrpc_pinger_remove_timeouts(void);

int ptlrpc_stop_pinger(void)
{
	struct l_wait_info lwi = { 0 };
	int rc = 0;

	if (!thread_is_init(&pinger_thread) &&
	    !thread_is_stopped(&pinger_thread))
		return -EALREADY;

	ptlrpc_pinger_remove_timeouts();
	thread_set_flags(&pinger_thread, SVC_STOPPING);
	wake_up(&pinger_thread.t_ctl_waitq);

	l_wait_event(pinger_thread.t_ctl_waitq,
		     thread_is_stopped(&pinger_thread), &lwi);

	return rc;
}

void ptlrpc_pinger_sending_on_import(struct obd_import *imp)
{
	ptlrpc_update_next_ping(imp, 0);
}
EXPORT_SYMBOL(ptlrpc_pinger_sending_on_import);

void ptlrpc_pinger_commit_expected(struct obd_import *imp)
{
	ptlrpc_update_next_ping(imp, 1);
	LASSERT(spin_is_locked(&imp->imp_lock));
	/*
	 * Avoid reading stale imp_connect_data.  When not sure if pings are
	 * expected or not on next connection, we assume they are not and force
	 * one anyway to guarantee the chance of updating
	 * imp_peer_committed_transno.
	 */
	if (imp->imp_state != LUSTRE_IMP_FULL ||
	    OCD_HAS_FLAG(&imp->imp_connect_data, PINGLESS))
		imp->imp_force_next_verify = 1;
}

int ptlrpc_pinger_add_import(struct obd_import *imp)
{
	if (!list_empty(&imp->imp_pinger_chain))
		return -EALREADY;

	mutex_lock(&pinger_mutex);
	CDEBUG(D_HA, "adding pingable import %s->%s\n",
	       imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd));
	/* if we add to pinger we want recovery on this import */
	imp->imp_obd->obd_no_recov = 0;
	ptlrpc_update_next_ping(imp, 0);
	/* XXX sort, blah blah */
	list_add_tail(&imp->imp_pinger_chain, &pinger_imports);
	class_import_get(imp);

	ptlrpc_pinger_wake_up();
	mutex_unlock(&pinger_mutex);

	return 0;
}
EXPORT_SYMBOL(ptlrpc_pinger_add_import);

int ptlrpc_pinger_del_import(struct obd_import *imp)
{
	if (list_empty(&imp->imp_pinger_chain))
		return -ENOENT;

	mutex_lock(&pinger_mutex);
	list_del_init(&imp->imp_pinger_chain);
	CDEBUG(D_HA, "removing pingable import %s->%s\n",
	       imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd));
	/* if we remove from pinger we don't want recovery on this import */
	imp->imp_obd->obd_no_recov = 1;
	class_import_put(imp);
	mutex_unlock(&pinger_mutex);
	return 0;
}
EXPORT_SYMBOL(ptlrpc_pinger_del_import);

/**
 * Register a timeout callback to the pinger list, and the callback will
 * be called when timeout happens.
 */
struct timeout_item* ptlrpc_new_timeout(int time, enum timeout_event event,
					timeout_cb_t cb, void *data)
{
	struct timeout_item *ti;

	OBD_ALLOC_PTR(ti);
	if (!ti)
		return(NULL);

	INIT_LIST_HEAD(&ti->ti_obd_list);
	INIT_LIST_HEAD(&ti->ti_chain);
	ti->ti_timeout = time;
	ti->ti_event = event;
	ti->ti_cb = cb;
	ti->ti_cb_data = data;

	return ti;
}

/**
 * Register timeout event on the the pinger thread.
 * Note: the timeout list is an sorted list with increased timeout value.
 */
static struct timeout_item*
ptlrpc_pinger_register_timeout(int time, enum timeout_event event,
			       timeout_cb_t cb, void *data)
{
	struct timeout_item *item, *tmp;

	LASSERT(mutex_is_locked(&pinger_mutex));

	list_for_each_entry(item, &timeout_list, ti_chain)
		if (item->ti_event == event)
			goto out;

	item = ptlrpc_new_timeout(time, event, cb, data);
	if (item) {
		list_for_each_entry_reverse(tmp, &timeout_list, ti_chain) {
			if (tmp->ti_timeout < time) {
				list_add(&item->ti_chain, &tmp->ti_chain);
				goto out;
			}
		}
		list_add(&item->ti_chain, &timeout_list);
	}
out:
	return item;
}

/* Add a client_obd to the timeout event list, when timeout(@time)
 * happens, the callback(@cb) will be called.
 */
int ptlrpc_add_timeout_client(int time, enum timeout_event event,
			      timeout_cb_t cb, void *data,
			      struct list_head *obd_list)
{
	struct timeout_item *ti;

	mutex_lock(&pinger_mutex);
	ti = ptlrpc_pinger_register_timeout(time, event, cb, data);
	if (!ti) {
		mutex_unlock(&pinger_mutex);
		return (-EINVAL);
	}
	list_add(obd_list, &ti->ti_obd_list);
	mutex_unlock(&pinger_mutex);
	return 0;
}
EXPORT_SYMBOL(ptlrpc_add_timeout_client);

int ptlrpc_del_timeout_client(struct list_head *obd_list,
			      enum timeout_event event)
{
	struct timeout_item *ti = NULL, *item;

	if (list_empty(obd_list))
		return 0;
	mutex_lock(&pinger_mutex);
	list_del_init(obd_list);
	/**
	 * If there are no obd attached to the timeout event
	 * list, remove this timeout event from the pinger
	 */
	list_for_each_entry(item, &timeout_list, ti_chain) {
		if (item->ti_event == event) {
			ti = item;
			break;
		}
	}
	LASSERTF(ti != NULL, "ti is NULL ! \n");
	if (list_empty(&ti->ti_obd_list)) {
		list_del(&ti->ti_chain);
		OBD_FREE_PTR(ti);
	}
	mutex_unlock(&pinger_mutex);
	return 0;
}
EXPORT_SYMBOL(ptlrpc_del_timeout_client);

int ptlrpc_pinger_remove_timeouts(void)
{
	struct timeout_item *item, *tmp;

	mutex_lock(&pinger_mutex);
	list_for_each_entry_safe(item, tmp, &timeout_list, ti_chain) {
		LASSERT(list_empty(&item->ti_obd_list));
		list_del(&item->ti_chain);
		OBD_FREE_PTR(item);
	}
	mutex_unlock(&pinger_mutex);
	return 0;
}

void ptlrpc_pinger_wake_up(void)
{
	thread_add_flags(&pinger_thread, SVC_EVENT);
	wake_up(&pinger_thread.t_ctl_waitq);
}

/* Ping evictor thread */
#define PET_READY     1
#define PET_TERMINATE 2

static int	       pet_refcount = 0;
static int	       pet_state;
static wait_queue_head_t       pet_waitq;
LIST_HEAD(pet_list);
static DEFINE_SPINLOCK(pet_lock);

int ping_evictor_wake(struct obd_export *exp)
{
	struct obd_device *obd;

	spin_lock(&pet_lock);
	if (pet_state != PET_READY) {
		/* eventually the new obd will call here again. */
		spin_unlock(&pet_lock);
		return 1;
	}

	obd = class_exp2obd(exp);
	if (list_empty(&obd->obd_evict_list)) {
		class_incref(obd, "evictor", obd);
		list_add(&obd->obd_evict_list, &pet_list);
	}
	spin_unlock(&pet_lock);

	wake_up(&pet_waitq);
	return 0;
}

static int ping_evictor_main(void *arg)
{
	struct obd_device *obd;
	struct obd_export *exp;
	struct l_wait_info lwi = { 0 };
	time_t expire_time;

	unshare_fs_struct();

	CDEBUG(D_HA, "Starting Ping Evictor\n");
	pet_state = PET_READY;
	while (1) {
		l_wait_event(pet_waitq, (!list_empty(&pet_list)) ||
			     (pet_state == PET_TERMINATE), &lwi);

		/* loop until all obd's will be removed */
		if ((pet_state == PET_TERMINATE) && list_empty(&pet_list))
			break;

		/* we only get here if pet_exp != NULL, and the end of this
		 * loop is the only place which sets it NULL again, so lock
		 * is not strictly necessary. */
		spin_lock(&pet_lock);
		obd = list_entry(pet_list.next, struct obd_device,
				     obd_evict_list);
		spin_unlock(&pet_lock);

		expire_time = cfs_time_current_sec() - PING_EVICT_TIMEOUT;

		CDEBUG(D_HA, "evicting all exports of obd %s older than %ld\n",
		       obd->obd_name, expire_time);

		/* Exports can't be deleted out of the list while we hold
		 * the obd lock (class_unlink_export), which means we can't
		 * lose the last ref on the export.  If they've already been
		 * removed from the list, we won't find them here. */
		spin_lock(&obd->obd_dev_lock);
		while (!list_empty(&obd->obd_exports_timed)) {
			exp = list_entry(obd->obd_exports_timed.next,
					     struct obd_export,
					     exp_obd_chain_timed);
			if (expire_time > exp->exp_last_request_time) {
				class_export_get(exp);
				spin_unlock(&obd->obd_dev_lock);
				LCONSOLE_WARN("%s: haven't heard from client %s"
					      " (at %s) in %ld seconds. I think"
					      " it's dead, and I am evicting"
					      " it. exp %p, cur %ld expire %ld"
					      " last %ld\n",
					      obd->obd_name,
					      obd_uuid2str(&exp->exp_client_uuid),
					      obd_export_nid2str(exp),
					      (long)(cfs_time_current_sec() -
						     exp->exp_last_request_time),
					      exp, (long)cfs_time_current_sec(),
					      (long)expire_time,
					      (long)exp->exp_last_request_time);
				CDEBUG(D_HA, "Last request was at %ld\n",
				       exp->exp_last_request_time);
				class_fail_export(exp);
				class_export_put(exp);
				spin_lock(&obd->obd_dev_lock);
			} else {
				/* List is sorted, so everyone below is ok */
				break;
			}
		}
		spin_unlock(&obd->obd_dev_lock);

		spin_lock(&pet_lock);
		list_del_init(&obd->obd_evict_list);
		spin_unlock(&pet_lock);

		class_decref(obd, "evictor", obd);
	}
	CDEBUG(D_HA, "Exiting Ping Evictor\n");

	return 0;
}

void ping_evictor_start(void)
{
	struct task_struct *task;

	if (++pet_refcount > 1)
		return;

	init_waitqueue_head(&pet_waitq);

	task = kthread_run(ping_evictor_main, NULL, "ll_evictor");
	if (IS_ERR(task)) {
		pet_refcount--;
		CERROR("Cannot start ping evictor thread: %ld\n",
			PTR_ERR(task));
	}
}
EXPORT_SYMBOL(ping_evictor_start);

void ping_evictor_stop(void)
{
	if (--pet_refcount > 0)
		return;

	pet_state = PET_TERMINATE;
	wake_up(&pet_waitq);
}
EXPORT_SYMBOL(ping_evictor_stop);